Skip to content

stripekit vs other tools

There are a few ways to get payments working quickly. They are not the same product.

Quick map

ApproachExamplesYour Stripe account?Who owns billing state?Runtime dependency?
Dev tool (stripekit)stripekitYes — yoursYour DB + StripeNo (CLI only)
Hosted control planeAutumnYes — under their APITheir serversYes
Merchant of recordPolar, Creem, DodoNo — they are the sellerTheir platformYes + fee %
Starter kit / patternShipFast, saas-starter, t3 recommendationsYesYour codeNo — but you still click the dashboard

When to pick stripekit

Use stripekit if you want:

  1. Plain Stripe as the source of truth (test and live are just two targets of the same config).
  2. Catalog in gitstripe.config.ts + plan/push, like Terraform for products/prices/webhook/portal.
  3. Code you own — webhook handler, checkout/portal routes, and the single-writer sync pattern scaffolded into your repo.
  4. No hosted billing middleman — nothing that can go down between you and Stripe, no revenue share for the tool.

Honest scope today: Next.js App Router scaffolding, subscriptions + one-time prices, Drizzle or KV sync. The reconciler itself is framework-agnostic; more scaffolds come later.

When to pick something else

  • Autumn (or similar) — you want entitlements, credits, usage meters, and a hosted API as the billing brain. Stripe is still underneath, but your app calls their control plane at runtime.
  • A merchant of record — you want them to handle tax/VAT, payouts, and being the legal seller. Different product: they replace your Stripe relationship (and take a cut).
  • A starter kit alone — fine for greenfield apps if you are happy creating products and webhooks by hand in the dashboard. stripekit is what you reach for when you want that catalog declarative and re-applicable to an existing app.

vs “just let an agent use Stripe’s MCP”

Stripe’s agent tooling is powerful and probabilistic: an agent can create products and wire webhooks, but each run can differ. stripekit is deterministic — the same config produces the same plan, tagged objects, and generated handlers. Agents still win: they call plan/push/check (or the MCP tools) instead of inventing money code.

Further reading

Released under the MIT License.